By Steven L. Ostrowski
Machine learning has recently gained a renewed interest as the technology powering it has become more widely available and accessible to organizations of all sizes. Applications using machine learning are being deployed in contexts and for purposes that were not even imaginable a few years ago.
This article is intended to start a conversation about how machine learning can be used as a force multiplier in public safety.
Machine learning can be used as a high-tech early warning system that can notify officers of potential risks, threats and areas for concern. (Photo/PoliceOne)
What is machine learning?
Shashank Gupta of the International Institute of Information Technology explains it this way: Imagine a dog you bought for protecting your home from intruders. To do that, you have to train your dog to distinguish between you or any unknown person. You do this by showing him unknown persons and training him to bark when it sees an unknown person. And then, you train him to not bark when it sees you.
Replace “dog” with a machine and “bark” with an alarm. This is machine learning.
You train a machine to learn pattern(s) present in data and then make decisions based on knowledge gathered.
Using machine learning in the public safety domain offers new opportunities for budget-strapped agencies to improve operational efficiencies and real-world outcomes.
Why do public safety agencies need a force multiplier?
Agencies are being asked to do more with less due to decreasing budgets, increasing responsibilities and a retiring workforce, while handling massive amounts of data to analyze from digital sources including video, social media, IoT sensors and the dark web.
“Sometimes, technology can get in the way of being effective and efficient in doing our jobs,” said Police Chief Gary J. Gacek of the Concord (N.C.) Police Department. “Machine learning is an emerging technology that shows promise.”
In one sense, machine learning can offer an extra pair of hands to do a repetitive task. In another sense, it can provide an extra pair of eyes and ears – ones that don’t tire from long overtime hours or get bored by monotonous tasks.
Police Chief Aaron Michael Raap of the Whitewater (Wis.) Police Department reminds us to “recognize that no type or amount of technology can replace human interaction, intuition, training and experience.” Further, Raap says he encourages “exploring and judiciously utilizing tools such as machine learning as a true force multiplier to more efficiently and effectively reduce crime, disorder and the perception of both.”
How can machine learning be used?
There are many ways machine learning can be used in public safety. Here is a brief description of four uses of machine learning.
1. Early warning for frontline officers
In the field, seconds count. Machine learning can provide officers warning of a threat not unlike how canaries were used by coal miners to detect methane or carbon monoxide in mines. The canary was a low-tech early warning system used to detect a gas before it became hazardous to humans.
Similarly, machine learning can be used as a high-tech early warning system that can notify officers of potential risks, threats and areas for concern. Machine learning models can be trained to monitor sensors and video cameras around an officer or a squad to detect irregular movements, anomalous activities, suspicious behaviors, dangerous objects and more.
2. Aid in solving cases
Machine learning can be used retroactively to understand what happened on a case. Likewise, it can be used to discover hidden relationships to aid in solving cases. Machine learning does this by “connecting the dots” between traditional data sources, video, photographs, narrative texts, forensics, shell casings and other pieces of evidence.
On most cases, there is too much data for investigators to dig through efficiently. That forces investigators to look at a subset of data using their intuition or a systemic process of prioritization. Machine learning can consider all the data and find relationships that might otherwise have been overlooked.
When a BOLO is issued, officers keep a lookout on the streets for persons of interest, suspect vehicles, license plates or missing persons. More recently, city video feeds are monitored. With new solutions come new problems. Imagine a person trying to monitor a video wall of a few hundred city video cameras to locate a person of interest. Don’t blink. Don’t be distracted by a text message from your daughter. Don’t use the washroom. Without machine learning automation, this is an overwhelming – if not impossible – task.
Once taught what to look for, a machine learning model can both monitor real-time video feeds and search through thousands of hours of previously recorded video, an amount well beyond what is humanly possible.
4. Crowd control
Machine learning models can be used to monitor social media where rallies and protests are now commonly communicated. Machine learning models can be taught to identify the formation of crowds in live video feeds, count the size of the crowd and track movement of the crowd.
Likewise, real-time video analysis can provide extra surveillance during planned mass gatherings like marathons, block parties and festivals. Machine learning models can be taught to pinpoint anomalous activity or suspicious behavior in a crowd. They can also be taught to identify objects such as weapons or unattended bags.
And the reverse is true: Machine learning can be used to identify sudden dispersion of a crowd, which can be an indicator of a threat.
Call to action
The scenarios described above are just a few ways that machine learning can be used in public safety as a force multiplier because it can process large volumes of information faster and with improved accuracy, plus lead to potential cost savings.
Retired Police Chief Edward A. Flynn of the Milwaukee Police Department says, “Police collect a great amount of data that has ramifications for policy formulation and context. How many? How often? Where? These are all questions not only about crime control but also the impact of police tactics on various communities.” He shares a few additional ideas on how machine learning might be used to assist public safety agencies:
Analyze calls for service (their distribution as well as outcome) for better prioritization decisions;
Control traffic flow (for example, to accelerate response times of emergency vehicles);
Detect reckless driving;
Identify communities requiring enhanced attention;
Evaluate the results of interventions and their impact on identified problems;
Accurately report what the public’s significant financial investments in public safety have accomplished.
As mentioned earlier, this article is intended to generate discussion on the uses of machine learning for public safety. Do you have additional ideas on how to use machine learning as a force multiplier? Please share your thoughts in the comments or drop me a message. Let’s keep the conversation going.
Flynn EA. Email interview. December 23, 2019.
Gacek GJ. LinkedIn interview. December 13, 2019.
Gupta S. “Learning the nuts and bolts of machine learning.” Quora, January 2, 2015.
Harpole J. LinkedIn interview. December 13, 2019.
Raap AM. LinkedIn interview. December 16, 2019.
About the author
Steven L. Ostrowski is a strategic business and technology advisor serving governments and Fortune 1000 companies with over two and a half decades of experience with SysLogic Inc and Accenture. He enjoys helping organizations make sense of technology that enables instead of entangles. Steven is honored to have had the opportunity to serve and learn from some of the nation’s finest including the FBI, ATF, WI DOJ, City of Milwaukee Police Department, Milwaukee County Sheriff, and Riley County Police Department, Manhattan. Follow Steven on Linkedin. Contact Steven L. Ostrowski.